Indian Railways Needs To Overhaul Train Scheduling Before Anything Else 

Much has been discussed about the merits and demerits of the Mumbai-Ahmedabad Bullet Train project. The Talgo train was also tried between Delhi and Mumbai to see how it may reduce travel time. Speeding up our railway system network-wide is a critical issue that has not been dwelt upon as even our current engines and rakes can run at least 30 kmph faster going by a very conservative estimate, without any major infrastructure upgrade beyond stringent checking of the railway track conditions.

However, there are only marginal gains to be made if one only focuses on the engine power and not the waiting time at railway signals. Therefore, a big bottleneck is the lack of any kind of optimisation in the scheduling of trains.

Consider a single line stretch between two stations, say A and B. There’s a train at A, heading towards B. And there’s a train at B, heading towards A. Whether the trains can travel at 50 kmph or 150 kmph, only one train can use the track at a time, and the other one will be waiting. Things can be bettered if in such a situation the single line stretch is replaced by a double line. However, if the scenario is expanded to resemble real world situations, every single step in the expansion throws up serious issues.

For example, if there is a double line system where there are junctions and intermediate points, either the trains are required to halt or use extra lines (of which one can have only so many). To understand this better, let us look at lock contention and semaphores in computer science and software enginnering parlance. Lock contention occurs when multiple databases require access to the same lock. The locks in question are held by multiple queued-up sessionsfor short spans of time and then released. This creates a single lane bridge situation. The problems are not obvious when the traffic is low, but once the traffic flow is high, things start to get bottlenecked. The semaphore itself was a form of a railway signal which indicated whether the track was already being used by another train, or ready for use.

The “greedy approach” to this problem is to have one or two high priority trains of the route which always get to run at their highest speed and other trains are made to halt for them, in case there is a contest for an available track. This is not a scalable solution as only one or two trains per route can get such a high “priority’ and when other trains get delayed, the delays can keep spilling over to tens or hundreds of trains in the Indian Railway system.

Therefore a well-planned, algorithmic scheduling of trains is a must. Railway scheduling is a complex academic problem in computer science and operations research. Throwing fancier hardware or newer engines on the system will be pointless unless the scheduling is correct.

Scheduling might even require all the trains to come to a halt for a day or two after which they restart using a fresh timetable. To illustrate the complexity and criticality of scheduling algorithms, here is a screenshot of some of the research papers one gets while making a web-search for this topic.

The famous Deccan Queen used to take just 2 hours 45 minutes from Mumbai to Pune, when it was launched in 1930. It now takes 3 hours 15 minutes despite the availability of far more powerful engines! This slowdown has happened gradually over the years. The reason is ad-hoc scheduling where existing trains suddenly need to have newer or longer halts, when new ones are introduced.

Coming to another famous heritage train, the Frontier Mail, between Mumbai and Kota (920 km) has only reduced about one and a half hours on an electrified stretch, while running with a very powerful WAP 7 electric engine. It used to take 15 hours 10 minutes in 1930 and it takes approximately 13 hours 40 minutes now.

It was re-christened to Golden Temple Mail about two decades ago but it still remains one of the faster trains on the Delhi-Mumbai route, after the Rajdhani Expresses. It arrived at Kota at 11 am in the 1930s and that timing holds true even today. So great is the challenge in re-scheduling trains that many trains from the pre-Independence era have similarly unchanged timings at various major halts. Even if one were to hypothetically transform the entire Mumbai-Delhi stretch into a bullet train corridor, only one or two trains would be able to run at bullet train speeds as the biggest bottleneck is optimised scheduling.

We have a similar case of the Punjab Mail which used to take 27 hours between Mumbai and Delhi in the early 1920s (via the Central Railway route). That was in an era of steam engines. Even today, with a fully electrified stretch and powerful electric engines, the train takes over 25 hours between Mumbai and New Delhi.

For similar reasons, the Tamil Nadu Express on the Delhi-Chennai route, takes 3 hours more today, than it did when it was launched over 40 years ago. The same is the case with the Karnataka Express (Delhi-Bangalore), Kerala Express (Delhi-Kerala) and Andhra Pradesh Express (Delhi-Secunderabad, now Telangana Express). All these trains which now average speeds of less than 70 kmph, ran on diesel power when launched four decades ago and currently run on powerful electric engines which can theoretically pull their rakes at over 120 kmph.

Apart from genuine track maintenance issues, it is the scheduling which is to blame.

A good description of these four historic south bound trains from Delhi, is available here.

Fancier engines will not make these trains any faster. They may run faster between two consecutive stations, but will end up waiting for a longer time as the departure time from a station cannot be changed without having a propagating effect on the timings of other trains.

Before spending a tonne of scarce resources for faster trains, India needs to wake up to this very real elephant in the room when it comes to our snail slow trains - the bottleneck is typically not the infrastructure. Indian railways needs to hire algorithmic engineers or network routing protocol designers or those with a background in Industrial Engineering and Operations Research. For now, this is not a problem which can be solved by throwing more resources. With our ad-hoc model of railway scheduling, if we start expanding the Mumbai-Ahmedabad Bullet Train corridor, we'll immediately run into similar problems and will not be network-wide average speeds befitting a high speed railway system.

Though the frequency of stops has certainly increased on many trains, the slowdown caused by stops is largely thanks to the unscheduled halts which occur 10 km before and after stations. This is again a scheduling issue. Given the advanced signalling systems and the powerful engines currently in use, versus the diesel and steam engines of the past, increased stoppages cannot be an acceptable excuse.

Here is the timetable of Frontier Mail in 1937, on the IRFCA website. It used to take 33 hours 25 minutes with a steam engine. It takes 32 hours 5 minutes even today despite being pulled by a powerful WAP 7 engine. This still carries the "superfast" tag and is still one of the faster trains on the Delhi - Mumbai route. It is important to note that the engines pulling the Rajdhani Express trains which take 4 hours less, are the same. Their faster speed is primarily because they are priority trains which almost always get to pass other trains in case there is a contest for a track. Again, a signalling and scheduling optimisation issue.

As things stand, the elephant in the room is completely oblivious to the mathematical and computational aspects of railway scheduling.